AMD Radeon RX Vega (Consumer version) Spotted in 3DMark FireStrike
Click here to post a comment for AMD Radeon RX Vega (Consumer version) Spotted in 3DMark FireStrike on our message forum
cryohellinc
Looks good!
JamesSneed
Running these on an Intel 5960x seems a bit suspicious.
Denial
If they sell it for $400 it will be competitive but I can't imagine AMD wanting that. It's a far bigger chip than the 1080, HBM2 is more expensive and the WC edition can't be priced that much more, so it's only going to eat into margins further. Next year they plan on refreshing it on 7nm with higher clocks (presumably) maybe that will be the crutch for the architecture, but so far it looks like the "hardware everything" that RTG wants (I guess because it requires less software developers) is a problem for efficiency and cost.
With Nvidia already shipping GV100 parts and consumer based variants most likely out in Q1 2018, I can't see AMD selling many of these, outside of miners maybe.
I'm hoping at lower clocks it's far more efficient for Ravenridge's sake.
The benchmarks were uploaded by "thegametechnician" who works in RTG's marketing. I don't know why he did it with a 5960x but yeah.
DeskStar
Hmmmmmmmmmm.........??
Not sure about this one. The card that is.... the pricing better be some crucial "shtuff" as these numbers need to be paired with something "tantalizing....!"
Toss3
Meh less than a 5% improvement over my overclocked GTX 980 Ti. The GTX 1080 at least overclocks pretty well, but if the power consumption of the FE is anything to go by this isn't going to be pretty.
rl66
why?
do you remember that Fury and Polaris best official presentation were on AMD's computer running intel CPU...
Ryzen is in "your box finalisation" 🙂 and TR is not lauched at all (and will end as Ryzen... BTW every cpu maker do the same... early consumer have higher price and get final result of their rig after a year).
So it look normal to get result on proofed high end (even from concurent) to see more or less what this GPU will have on own high end when ready.
riot83
So my 1080 @ 2050 mhz has 2k higher graphics score. I'm sure they'll catch up to it with driver updates.
http://www.3dmark.com/fs/12996779
Edit: I just wonder why the 5960X's(physics) score is so low. My 5920k is higher by 2k.
coth
Power consumption is disastrous. What's more it appears to be throttling from 84 sec on air.
https://www.pcper.com/image/view/84106
Exascale
Wrong card. Wrong thread.
Loophole35
That 5960x is likely running at default settings. A test rig such as that would not be overclocked.
fantaskarsef
Doesn't look too bad tbh. Still depends on pricing and availability at launch...
Evildead666
Yet again, where do you have the Info that Nvidia is "shipping" GV100 consumer parts ?
Tesla V100's are not consumer cards. They are not GV100's.
There will never be a consumer GV100.
Maybe a GV102.....down the line.
Singleton99
As i'm without Graphics card atm ive decided to wait and see how the AMD Vega going to look before pulling the trigger on the 1080 ti .
Now we just need some gaming benchmarks popping up on the web to tell more of the story on Vega so i can decide . 🙂
coth
Pretty much same GPU. So it is representative for RX Vega as well.
Denial
Consumer variants of Volta man, not GV100, obviously. I responded to you in the other thread with a link. And yeah, they are definitely GV100, they have 4 SM's disabled, but that doesn't make them "not GV100". P100 also had disabled SM's, doesn't make it "not GP100".
The point is Volta architecture is out and in people's hands. It's only a matter of time before the consumer variants are out and on the market. Which makes the shelf life for Vega limited.
EdensRemorse
Hard to believe this is the full-fat chip tbh
Polaris 20 gets around 2250 points in Firestrike per tflop. Fiji XT sits around 1860 per tflop. The score for Vega is a mere 1675 points per tflop assuming 64 CUs. This regression in what amounts to a measure of IPC from Polaris to Vega is astonishing. If it were a 56 CU chip being tested here, we'd be looking at 1900 points per tflop, which isn't great, but at least more in line with Fiji XT!?!
I was really hoping to see AMD get over the unbelievably bandwidth starved nature of Vega FE by enabling hardware features and improving software optimization, but apparently they're not there yet. Still, I expect full-fat Vega RX to come out swinging with 28-29K in Firestrike at stock clocks, with memory OCs clearing 30k - call me an optimist...
$399 for 56 CUs, $549 for 64 CUs
Singleton99
After having a look around the web , theres three different versions, so this benchmark may not be the full fat Vega
RX Vega Core =1070
RX Vega Eclipse =1080
RX Vega Nova= 1080ti
A M D BugBear
Awaiting for firestrike ultra scores, screw all others, we want to see what the card card can do when its fully utilized instead of lower resolution/settings = more favor to the cpu, less/burden towards the gpu, we want to see the total opposite, this way we can see clearly how powerful the card is, stress the card alot and see what it can do,instead of being bottlenecked from cpu because of super lower resolution, up those resolution and settings, put all favor to the gpu, this way we can see everything what the card can do, leave the lowest favor to the cpu.
Can't wait for ultra scores, all others(Firestrike bench speaking), not interested, card is built for much higher then 1080p, even with filters turned on.
Reviews, can't wait.
Robbo9999
I don't think Firestrike is CPU limited by any stretch, my GTX 1070 reaching 21000 GPU points is always at 100% GPU utilisation during the tests. I'd imagine these AMD cards are also at 100% GPU utilisation during Firestrike too - so I think these Firestrike scores we see in this article are valid & comparative.
A M D BugBear
Ok, let me take that back, I am so used to insane res and settings, Once your into these settings, Can't go back, Its pretty bad, lol, in a good way.
